In Cringe Video, OpenAI CTO Says She Doesn’t Know Where Sora’s Training Data Came From
In Cringe Video, OpenAI CTO Says She Doesn’t Know Where Sora’s Training Data Came From
Wondering what data OpenAI used to train its buzzy new text-to-video AI? OpenAI CTO Mira Murati seems to be wondering, too.
I hope this is gonna become a new meme template
Coffeezilla had a video in his void where he plays this back a few times. It's hilarious seeing the guilt without stating it.
They know what they fed the thing. Not backing up their own training data would be insane. They are not insane, just thieves
Everyone says this but the truth is copyright law has been unfit for purpose for well over 30 years now. And the lords were written no one expected something like the internet to ever come along and they certainly didn't expect something like AI. We can't just keep applying the same old copyright laws to new situations when they already don't work.
I'm sure they did illegally obtain the work but is that necessarily a bad thing? For example they're not actually making that content available to anyone so if I pirate a movie and then only I watch it, I don't think anyone would really think I should be arrested for that, so why is it unacceptable for them but fine for me?
That is a bad thing if they want to be exempt from the law because they are doing a big, very important thing, and we shouldn't.
The copyright laws are shit, but applying them selectively is orders of magnitude worse.
if I pirate a movie and then only I watch it, I don't think anyone would really think I should be arrested for that, so why is it unacceptable for them but fine for me?
Because it's more analogous to watching a video being broadcasted outdoors in the public, or looking at a mural someone painted on a wall, and letting it inform your creative works going forward. Not even recording it, just looking at it.
As far as we know, they never pirated anything. What we do know is it was trained on data that literally anybody can go out and look at yourself and have it inform your own work. If they're out here torrenting a bunch of movies they don't own or aren't licencing, then the argument against them has merit. But until then, I think all of this is a bunch of AI hysteria over some shit humans have been doing since the first human created a thing.
That's really not how it works though, it's a web crawler they're not going to download the whole internet
And a reason they don't is it would actually potentially be copywrite infringement in some cases where as what they do legally isn't (no matter how much people wish the law was set based on their emotions)
Gee, seems like something a CTO would know. I'm sure she's not just lying, right?
And on the other hand it is a very obvious question to expect. If you have something hide how on the world are you not prepared for this question !? 🤡
It's a question that is based on a purposeful misunderstanding of the technology, it's like expecting a bee keeper to know each bees name and bedtime. Really it's like asking a bricklayer where each brick came from in the pile, He can tell you the batch but not going to know this brick came from the forth row of the sixth pallet, two from the left. There is no reason to remember that it's not important to anyone.
The don't log it because it would take huge amounts of resources and gain nothing.
What?
Compiling quality datasets is enormously challenging and labour intensive. OpenAI absolutely knows the provenance of the data they train on as it's part of their secret sauce. And there's no damn way their CTO won't have a broad strokes understanding of the origins of those datasets.
[Citation needed]
To be fair, these datasets are one of their biggest competitive edge. But saying in to interviewer "I cannot tell you", is not very nice, so you can take the americal politician approach and say "I don't know/remember" which you cannot ever be hold accountable for.
There is no way in hell it isn’t copyrighted material.
Every video ever created is copyrighted.
The question is — do they need a license? Time will tell. This is obviously going to court.
Don't downvote this guy. He's mostly right. Creative works have copyright protections from the moment they are created. The relevant question is indeed if they have the relevant permissions for their use, not wether it had protections in the first place.
Maybe some surveillance camera footage is not sufficiently creative to get protections, but that's hardly going to be good for machine reinforcement learning.
There are definitely non copyrighted videos! Both old videos (all still black and white I think) and also things released into the public domain by copyright holders.
But for sure that's a very small subset of videos.
If I were the reporter my next question would be:
"Do you feel that not knowing the most basic things about your product reflects on your competence as CTO?"
Hilarious, but if the reporter asked this they would find it harder to get invites to events. Which is a problem for journalists. Unless your very well regarded for your journalism, you can't push powerful people without risking your career.
Also about this line:
Others, meanwhile, jumped to Murati's defense, arguing that if you've ever published anything to the internet, you should be perfectly fine with AI companies gobbling it up.
No I am not fine. When I wrote that stuff and those researches in old phpbb forums I did not do it with the knowledge of a future machine learning system eating it up without my consent. I never gave consent for that despite it being publicly available, because this would be a designation of use that wouldn't exist back than. Many other things are also publicly available, but some a re copyrighted, on the same basis: you can publish and share content upon conditions that are defined by the creator of the content. What's that, when I use zlibrary I am evil for pirating content but openai can do it just fine due to their huge wallets? Guess what, this will eventually creating a crisis of trust, a tragedy of the commons if you will when enough ai generated content will build the bulk of your future Internet search! Do we even want this?
I almost want to believe they legitimately do not know nor care they‘re committing a gigantic data and labour heist but the truth is they know exactly what they‘re doing and they rub it under our noses.
Of course they know what they’re doing. Everybody knows this, how could they be the only ones that don’t?
Yeah, the fact that AI progress just relies on "we will make so much money that no lawsuit will consequently alter our growth" is really infuriating. The fact that general audience apparently doesn't care is even more infuriating.
Look guys! I'm stealing from Tolkien!
This tellls you so much what kind of company OpenAI is
An Intelligence piracy company?
Half open or half close?
It also tells us how hypocritical we all are since absolutely every single one of us would make the same decisions they have if we were in their shoes. This shit was one bajillion percent inevitable; we are in a river and have been since we tilled soil with a plough in the Nile valley millennia ago.
Did they intentionally chose a picture where she looks like she's morphing into Elon?
I was thinking mads mikkelssen
Well after just finishing Death Stranding, I can’t unsee that.
I suspect so. It is a very slanted article.
CTO should definitely know this.
They do know this. They're avoiding any legal exposure by being vague.
I feel like at their scale, if there's going to be a figure head marketable CTO, it's going to be this company. If not, you're right, and she's lying lol
Of course she knows it. She just doesn't want to get sued.
Funny she didn't talked it out with lawyers before that. That's a bad way to answer that.
Or she talked and the lawyers told her to pretend ignorance.
It probably means that they don’t scrape and preprocess training data in house. She knows they get it from a garden variety of underpaid contractors, but she doesn’t know the specific data sources beyond the stipulations of the contract (“publicly available or licensed”), and she probably doesn’t even know that for certain.
Then wipe it out and start again once you have where your data is coming from sorted out. Are we acting like you having built datacenter pack full of NVIDIA processors just for this sort of retraining? They are choosing to build AI without proper sourcing, that's not an AI limitation.
REPORTER: Where does your data come from?
CTO: Bitch, are you trying to get me sued?
So plagiarism?
I don't think so. They aren't reproducing the content.
I think the equivalent is you reading this article, then answering questions about it.
Idk why this is such an unpopular opinion. I don't need permission from an author to talk about their book, or permission from a singer to parody their song. I've never heard any good arguments for why it's a crime to automate these things.
I mean hell, we have an LLM bot in this comment section that took the article and spat 27% of it back out verbatim, yet nobody is pissing and moaning about it "stealing" the article.
This is the best summary I could come up with:
Mira Murati, OpenAI's longtime chief technology officer, sat down with The Wall Street Journal's Joanna Stern this week to discuss Sora, the company's forthcoming video-generating AI.
It's a bad look all around for OpenAI, which has drawn wide controversy — not to mention multiple copyright lawsuits, including one from The New York Times — for its data-scraping practices.
After the interview, Murati reportedly confirmed to the WSJ that Shutterstock videos were indeed included in Sora's training set.
But when you consider the vastness of video content across the web, any clips available to OpenAI through Shutterstock are likely only a small drop in the Sora training data pond.
Others, meanwhile, jumped to Murati's defense, arguing that if you've ever published anything to the internet, you should be perfectly fine with AI companies gobbling it up.
Whether Murati was keeping things close to the vest to avoid more copyright litigation or simply just didn't know the answer, people have good reason to wonder where AI data — be it "publicly available and licensed" or not — is coming from.
The original article contains 667 words, the summary contains 178 words. Saved 73%. I'm a bot and I'm open source!
Funny how we have all this pissing and moaning about stealing, yet nobody ever complains about this bot actually lifting entire articles and spitting them back out without ads or fluff. I guess it's different when you find it useful, huh?
I like the bot, but I mean y'all wanna talk about copyright violations? The argument against this bot is a hell of a lot more solid than just using data for training.
So my work uses ChatGPT as well as all the other flavours. It's getting really hard to stay quiet on all the moral quandaries being raised on how these companies are training their AI data.
I understand we all feel like we are on a speeding train that can't be stopped or even slowed down but this shit ain't right. We need to really start forcing businesses to have moral compass.
I spot aot of people GPT-eing their way through personale notes and researches. Whereas you used to see Evernote, office, word, note taking app you see a lot of gpt now. I feel weird about it.
this is why code AND cloud services shouldn't be copyrightable or licensable without some kind of transparency legislation to ensure people are honest. Either forced open source or some kind of code review submission to a government authority that can be unsealed in legal disputes.
Obviously nobody fully knows where so much training data come from. They used Web scraping tool like there's no tomorrow before, with that amount if informations you can't tell where all the training material come from. Which doesn't mean that the tool is unreliable, but that we don't truly why it's that good, unless you can somehow access all the layers of the digital brains operating these machines; that isn't doable in closed source model so we can only speculate. This is what is called a black box and we use this because we trust the output enough to do it. Knowing in details the process behind each query would thus be taxing. Anyway...I'm starting to see more and more ai generated content, YouTube is slowly but surely losing significance and importance as I don't search informations there any longer, ai being one of the reasons for this.
what's wrong with her face?
she grimaced?
They use awkward stills to generate clicks
It's annoying and distracting, just like the headline.
It's an AI.
I have a feeling that the training material involves cheese pizza...
LLM is just another iteration of Search. Search engines do the same thing. Do we outlaw search engines?
SoRA is a generative video model, not exactly a large language model.
But to answer your question: if all LLMs did was redirect you to where the content was hosted, then it would be a search engine. But instead they reproduce what someone else was hosting, which may include copyrighted material. So they’re fundamentally different from a simple search engine. They don’t direct you to the source, they reproduce a facsimile of the source material without acknowledging or directing you to it. SoRA is similar. It produces video content, but it doesn’t redirect you to finding similar video content that it is reproducing from. And we can argue about how close something needs to be to an existing artwork to count as a reproduction, but I think for AI models we should enforce citation models.
How does a search engine know where to point you? It injests all that data and processes it 'locally' on the search engines systems using algorithms to organize the data for search. It's effectively the same dataset.
LLM is absolutely another iteration of Search, with natural language ouput for the same input data. Are you advocating against search engine data injest as not fair use and copyright violations as well?
You equate LLM to Intelligence which it is not. It is algorithmic search interation with natural language responses, but that doesn't sound as cool as AI. It's neat, it's useful, and yes, it should cite the sourcing details (upon request), but it's not (yet?) a real intelligence and is equal to search in terms of fair use and copyright arguments.
I feel conflicted about the whole thing. Technically it's a model. I don't feel that people should be able to sue me as a scientist for making a model based on publicly available data. I myself am merely trying to use the model itself to explain stuff about the world. But OpenAI are also selling access to the outputs of the model, that can very closely approximate the intellectual property of people. Also, most of the training data was accessed via scraping and other gray market methods that were often explicitly violating the TOU of the various places they scraped from. So it all is very difficult to sort through ethically.
Don't know why you are down voted it's a good question.
As a matter of fact it almost happened for search engines in France. Newspaper's argued that snippets were leading people to not go into their ad infested sites thus losing them revenue.
https://techcrunch.com/2020/04/09/frances-competition-watchdog-orders-google-to-pay-for-news-reuse/
Any company CEO does not know shit that goes on in the dev department so her answer does not surprise me, ask the Devs or the team leader in charge of the project. The CEO is only there to make sure the company makes money as he and the share holders only care about money!
She's CTO not CEO. She absolutely should know the answer.
She knows the answer. She doesn’t the legal status of the answer, so she blanks. Been there before, I’ve got some sympathy for being in the limelight and being asked a tough question.
As my media trainer said, if you aren’t willing to discuss a subject, make it a condition of the interview. Once the camera rolls, declining to answer seems incredibly suspect.
She should but she does not as I mention in another post anyone at team leader or above in all the companies that I work so far bearly had any technical skill and didn't have any idea about this shit, only some bits and pieces that they got through some documentation that the dev team made. They had some vague idea of how our infrastructure works but that about it.
Chief Technology Officer, not CEO
So you mean another person that has no idea because is higher up on the chain of command that all he/she cares about is how to make more money ? Seriously in any company I worked untill not everyone at the level of management or above had mostly no idea about this shit and most of them I have no idea how they got in those positions as they have close to 0 technical skill! And the speeches that those people do are made by people that again are not part of the infrastructure or development team. I do find this disturbing as hell but at this point it's also what I expect to happend as I only seen this shit.
Watching a video or reading an article by a human isn't copyright infringement, why then if an "AI" do it then it is? I believe the copyright infringement it's made by the prompt so by the user not the tool.
This is what people fundamentally don't understand about intelligence, artificial or otherwise. People feel like their intelligence is 100% "theirs". While I certainly would advocate that a person owns their intelligence, It didn't spawn from nothing.
You're standing on the shoulders of everyone that came before you. You take a prehistoric man or an alien that hasn't had any of the same experiences you've had, they won't be able to function in this world. It's not because they are any dumber than you. It's because you absorbed the hive mind of the society you live in. Everyone's racing to slap their brand on stuff to copyright it to get ahead and carve out their space.
"No you can't tell that story, It's mine." "That art is so derivative."
But copyright was only meant to protect something for a short period in order to monetize it; to adapt the value of knowledge for our capital market. Our world can't grow if all knowledge is owned forever and isn't able to be used when even THINKING about new ideas.
ANY VERSION OF INTELLIGENCE YOU WOULD WANT TO INTERACT WITH MUST CONSUME OUR KNOWLEDGE AND PRODUCE TRANSFORMATIONS OF IT.
That's all you do.
Imagine how useless someone would be who'd never interacted with anything copyrighted, patented, or trademarked.
That's not a very agreeable take. Just get rid of patents and copyrights altogether and your point dissolves itself into nothing. The core difference being derivative works by humans can respect the right to privacy of original creators.
Deep learning bullshit software however will just regurgitate creator's contents, sometimes unrecognizable, but sometimes outright steal their likeness or individual style to create content that may be associated with the original creators.
what you are in effect doing, is likening learning from the ideas of others to a deep learning "AI" using images for creating revenge porn, to give a drastic example.
What does this human is going to do with this reading ? Are they going to produce something by using part of this book or this article ?
If yes, that's copyright infringement.
When a school professor "prompts" you to write an essay and you, the "tool" go consume copyrighted material and plagiarize it in the production of your essay is the infringement made by the professor?